Some features:
The package uses the grid plotting system (just like ggplot2).
Node scaling
Node shapes
Edge curvature
Edge type of line
The distribution of \(\mathbf{Y}\) can be parameterized in the form
\[ \Pr\left(\mathbf{Y}=\mathbf{y}|\theta, \mathcal{Y}\right) = \frac{\exp{\theta^{\mbox{T}}\mathbf{g}(\mathbf{y})}}{\kappa\left(\theta, \mathcal{Y}\right)},\quad\mathbf{y}\in\mathcal{Y} \tag{1} \]
Where \(\theta\in\Omega\subset\mathbb{R}^q\) is the vector of model coefficients and \(\mathbf{g}(\mathbf{y})\) is a q-vector of statistics based on the adjacency matrix \(\mathbf{y}\).
Model (1) may be expanded by replacing \(\mathbf{g}(\mathbf{y})\) with \(\mathbf{g}(\mathbf{y}, \mathbf{X})\) to allow for additional covariate information \(\mathbf{X}\) about the network. The denominator,
\[ \kappa\left(\theta,\mathcal{Y}\right) = \sum_{\mathbf{z}\in\mathcal{Y}}\exp{\theta^{\mbox{T}}\mathbf{g}(\mathbf{z})} \]
Is the normalizing factor that ensures that equation (1) is a legitimate probability distribution.
Even after fixing \(\mathcal{Y}\) to be all the networks that have size \(n\), the size of \(\mathcal{Y}\) makes this type of models hard to estimate as there are \(N = 2^{n(n-1)}\) possible networks!
An Extension of the ergm (regular size fitting via simulation) package
Uses exact statistics for fitting small networks (3 to 6 nodes).
Will be designed mostly to be ran with multiple networks simulatenously (so we recover the asymptotic properties of the MLE estimators)
Work in progress…
netplot: https://github.com/USCCANA/netplot
Applied SNA with R: https://gvegayon.github.io/appliedsnar/
Little ERGMs: https://github.com/USCCANA/social-smarts/
Twitter: @gvegayon
email: vegayon@usc.edu